BYU

Abstract by Nathan Anderson

Personal Infomation


Presenter's Name

Nathan Anderson

Co-Presenters

Jonathan Skaggs
Trevin Avery
Josh Nelson

Degree Level

Undergraduate

Co-Authors

None

Abstract Infomation


Department

Computer Science

Faculty Advisor

Jacob Crandall

Title

Composable AI

Abstract

Most current approaches for developing AI are expensive, time-consuming, and require large amounts of training data. Even when these approaches are successful, the computer generally cannot transfer acquired knowledge to new domains. Composable AI aims to allow the transfer of learned skills to novel situations. In Composable AI, a group of users (including both programmers and non-technical users) will teach a computer to execute simple tasks, store those basic skills as components, and recombine these components to solve novel tasks. Each individual component is so basic that it is not likely to be viewed as intelligent. However, these components can be joined together to create a system that appears intelligent because it is capable of (1) performing complex cognitive tasks, (2) easily adapting its capabilities to other domains, (3) engaging in continual improvement, and (4) conversing with people about its behavior. Our research team has been creating a Composable AI programming framework to test these hypotheses.